CN105045901B - The method for pushing and device of search key - Google Patents
The method for pushing and device of search key Download PDFInfo
- Publication number
- CN105045901B CN105045901B CN201510475976.9A CN201510475976A CN105045901B CN 105045901 B CN105045901 B CN 105045901B CN 201510475976 A CN201510475976 A CN 201510475976A CN 105045901 B CN105045901 B CN 105045901B
- Authority
- CN
- China
- Prior art keywords
- search
- user
- keyword
- scene
- described search
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/951—Indexing; Web crawling techniques
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention discloses a kind of method for pushing of search key and devices.The described method includes: obtaining search key relevant to the reading context and scene of user from least one data source；Its corresponding clicking rate is estimated to described search keyword, and described search keyword is ranked up according to the clicking rate that estimation obtains；Described search keyword is pushed to the desktop terminal of user according to the order of the sequence.The method for pushing and device of search key provided in an embodiment of the present invention are that different users customizes different search keys, realize personalized search key push.
Description
Technical field
The present embodiments relate to search engine technique field more particularly to the method for pushing and dress of a kind of search key
It sets.
Background technique
Along with the development of internet, more and more users begin to use search engine.With the use of search engine
The promotion of rate, search engine start to release some supplementary means to improve search efficiency.In these supplementary means, search is crucial
The automatic recommendation of word is wherein very important one kind.The automatic recommendation of so-called search key, i.e. search box when the user clicks,
Or when inputting some keyword, a series of relevant passes of keyword that may be inputted to user can occur in result of page searching
Keyword.If user clicks these relevant keywords, user can further be searched for.
The excavation mode of existing search key is a kind of passively mode.Specifically, some search can generally be chosen
Index holds up the heat that developer rule of thumb obtains and searches word, these heat are then searched word and are pushed to user.This push mode
Defect is, seldom considers the individual demand of user, and often there is no what areas for the search key pushed to different user
Not.
Summary of the invention
In view of the above technical problems, the embodiment of the invention provides a kind of method for pushing of search key and device, from
And be that different users customizes different search keys, realize personalized search key push.
In a first aspect, the embodiment of the invention provides a kind of method for pushing of search key, which comprises
Search key relevant to the reading context and scene of user is obtained from least one data source；
Its corresponding clicking rate is estimated to described search keyword, and described search is closed according to the clicking rate that estimation obtains
Keyword is ranked up；
Described search keyword is pushed to the desktop terminal of user according to the order of the sequence.
Second aspect, the embodiment of the invention also provides a kind of driving means of search key, described device includes:
Module is obtained, is closed for obtaining search relevant to the reading context and scene of user from least one data source
Keyword；
Sorting module, for estimating its corresponding clicking rate, and the click obtained according to estimation to described search keyword
Rate is ranked up described search keyword；
Pushing module, for pushing described search keyword to the desktop terminal of user according to the order of the sequence.
The method for pushing and device of search key provided in an embodiment of the present invention from least one data source by obtaining
Search key relevant to the reading context and scene of user estimates its corresponding clicking rate to described search keyword,
And described search keyword is ranked up according to estimation obtained clicking rate, and according to the order of the sequence to user's
Desktop terminal pushes described search keyword, customizes different search keys for different users, realizes personalized search
The push of rope keyword.
Detailed description of the invention
By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, of the invention other
Feature, objects and advantages will become more apparent upon:
Fig. 1 is the flow chart of the method for pushing for the search key that first embodiment of the invention provides；
Fig. 2 is the flow chart of the push step in the method for pushing for the search key that second embodiment of the invention provides；
Fig. 3 is the display interface background for the search key that second embodiment of the invention provides；
Fig. 4 A is the display renderings for the search key that second embodiment of the invention provides；
Fig. 4 B is that the heat for the search key that second embodiment of the invention provides searches the display renderings of picture；
Fig. 4 C is that the heat for the search key that second embodiment of the invention provides searches the display renderings of voice；
Fig. 5 is the flow chart of the method for pushing for the search key that third embodiment of the invention provides；
Fig. 6 is the flow chart of obtaining step in the method for pushing for the search key that fourth embodiment of the invention provides；
Fig. 7 is the flow chart of the method for pushing for the search key that fifth embodiment of the invention provides；
Fig. 8 is the flow chart of obtaining step in a kind of preferred embodiment of fifth embodiment of the invention；
Fig. 9 is the flow chart of obtaining step in another preferred embodiment of fifth embodiment of the invention；
Figure 10 is the schematic illustration of Random Walk Algorithm in another preferred embodiment of fifth embodiment of the invention；
Figure 11 is the flow chart of obtaining step in the method for pushing for the search key that sixth embodiment of the invention provides；
Figure 12 is the flow chart of obtaining step in the method for pushing for the search key that seventh embodiment of the invention provides；
Figure 13 is the flow chart of sequence step in the method for pushing for the search key that eighth embodiment of the invention provides；
Figure 14 is the flow chart of sequence step in the method for pushing for the search key that ninth embodiment of the invention provides；
Figure 15 is the structure chart of the driving means for the search key that tenth embodiment of the invention provides.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining the present invention rather than limiting the invention.It also should be noted that in order to just
Only the parts related to the present invention are shown in description, attached drawing rather than entire infrastructure.
First embodiment
Present embodiments provide a kind of technical solution of the method for pushing of search key.The push of described search keyword
Method is executed by the driving means of search key, also, the driving means of described search keyword is generally integrated in network side
Search engine server etc. calculate in equipment.
Referring to Fig. 1, the method for pushing of described search keyword includes:
S11 obtains search key relevant to the reading context and scene of user from least one data source.
In the present embodiment, for obtain the data source of search key can have it is multiple.For example, can be by user's history
Log, internet web page, any one or their combination in search log in real time are as obtaining described search keyword
Data source.The user's history log refers to and uses the journal file of the historical operation of search engine for recording user.Tool
Body, the search term inputted when record has user before a period of time using search engine in the user's history log,
The information such as the web page interlinkage checked are clicked on result of page searching.
The internet web page refers on the internet, the Webpage that the public is able to access that.Further, described mutual
Should include user in intranet web can be from the content-data for wherein getting search key.This general feelings of content-data
It is to be presented to the public's in the form of text under condition.
The real-time search log refers to the journal file recorded in real time to user using the movement of search engine.Institute
The feature for stating search log maximum in real time is the timeliness of itself.That is, the real-time search log is able to reflect
User uses the real-time action of search engine.
The reading context of user refers to the context for the content that user is read by internet.For example, the reading of user
Content is if it is an internet web page, then the reading context can be user before or after reading this webpage
The web page address clicked.
The scene includes page scene and user's scene.Page scene may include the Type of website, page properties, the page
The parameters such as content.User's scene may include the parameters such as the access equipment of user, source place.
The search key that will be pushed to user is obtained based on the reading context of user and scenario parameters, is got
Search key will because of user itself reading habit and scenario parameters and change.In this way, for different use
Family, the search key that search engine is got will be different.Even the same user, in the period of different, search engine
The search key got also will be different.That is, the search key got can fully consider readding for user itself
Habit and scene are read, thus to realize that personalized search key push provides premise.
S12 estimates its corresponding clicking rate to described search keyword, and is searched according to the clicking rate that estimation obtains to described
Rope keyword is ranked up.
It is understood that illustrate that the search key is more effective if the clicking rate of a search key is high,
User is more willing to carry out searching for Internet using the search key.Therefore, when pushing search key, it should select as far as possible
Clicking rate search key in the top, is pushed.
In order to push more efficiently search key to user, clicking rate estimation is carried out to described search keyword, and
The clicking rate obtained according to estimation is ranked up described search keyword.Specifically, can be according to the clicking rate that formula (1) provides
Estimation model estimates the clicking rate of different search keys.
In formula (1), CTRtemplateIndicate the clicking rate in advertisement position stage, CTRluIndicate the click in link unit stage
Rate.xiIt indicates in advertisement position stage described search keyword, the feature vector element of corresponding page scene, user's scene, wiTable
Show the corresponding weight vectors element of described eigenvector element.xjIt indicates in connection unit stage described search keyword, correspondence
Page scene, user's scene feature vector element, wjIt indicates corresponding in link unit stage described eigenvector element
Weight vectors element.Wherein, described eigenvector element can be and described to search plain keyword co-occurrence in same data source general
The higher word of rate.The weight vectors element is then weighted value ginseng of the described eigenvector element in clicking rate estimation model
Number.
As can be seen that being divided into the clicking rate estimation in advertisement position stage in the clicking rate estimation procedure based on formula (1) model
Estimate with the clicking rate in link unit stage.In the advertisement position stage, the clicking rate parameter of estimation is the click of entire advertisement position
Rate；And in the link unit stage, the clicking rate parameter of estimation is a specific click of the search key on the advertisement position
Rate.
S13 pushes described search keyword to the desktop terminal of user according to the order of the sequence.
Preferably, can be related to showing the boundary of described search keyword according to the screen width value range of desktop terminal
Face layout, further according to the interface layout, shows described search keyword to desktop terminal.
It is further preferred that can be on the interface each search key design its it is corresponding show region, and
It is equipped with corresponding background colour for its region that shows, to reach the beauty of visual effect.
Moreover, when pushing described search keyword searching for recommendation can be adjusted by the adjustment to presentation time parameter
Rope keyword shows order, to guarantee the timeliness for being pushed to the search key of user.
The present embodiment is closed by obtaining search relevant to the reading context and scene of user from least one data source
Keyword estimates its corresponding clicking rate to described search keyword, and the clicking rate obtained according to estimation is to described search key
Word is ranked up, and pushes described search keyword to the desktop terminal of user according to the order of the sequence, thus for not
Same user customizes different search keys, realizes personalized search key push.
Second embodiment
The present embodiment further provides the method for pushing of search key based on the above embodiment of the present invention
A kind of technical solution of middle push step.In the technical scheme, it is pushed away according to the order of the sequence to the desktop terminal of user
Sending described search keyword includes: the screen width according to desktop terminal, designs the interface layout of each search key；According to institute
Interface layout is stated, to user shows described search keyword, the corresponding heat of described search keyword searches picture and/or described search
The corresponding heat of keyword searches voice.
Referring to fig. 2, include: to the desktop terminal of user push described search keyword according to the order of the sequence
S21 designs the interface layout of each search key according to the screen width of desktop terminal.
Fig. 3 illustratively shows a kind of interface layout at the displaying interface of described search keyword.Referring to Fig. 3, in institute
It states and shows on interface, the display area of different search keys is marked with the color lump of different colors.For clicking rate height, or
Text itself has the search key of certain length, the biggish color lump of usable floor area；And it is low for clicking rate, or text itself
The lesser search key of word length, the lesser color lump of usable floor area.
Can according to show be season difference, use different colours tone color lump complete interface layout.For example, in the winter
Interface layout can be carried out using the color lump of cool tone when season, and interface cloth can be carried out using warm-toned color lump in summer
Office.
S22 shows described search keyword, the corresponding heat of described search keyword to user according to the interface layout
It searches picture and/or the corresponding heat of described search keyword searches voice.
It, can be in pair of the interface layout after screen width according to desktop terminal completes the interface layout design
Region is answered to show described search keyword.Fig. 4 A, Fig. 4 B and Fig. 4 C respectively illustrate the corresponding position exhibition in the display interface
Existing described search keyword, the corresponding heat of described search keyword searches picture and the corresponding heat of described search keyword searches voice
Display renderings.
The present embodiment designs the interface layout of each search key, Yi Jigen by the screen width according to desktop terminal
According to the interface layout, to user shows described search keyword, the corresponding heat of described search keyword searches picture and/or described
The corresponding heat of search key searches voice, to not only realize the content personalization of search key push, but also realizes
The personalization at search key push interface.
3rd embodiment
The present embodiment further provides the method for pushing of search key based on the above embodiment of the present invention
A kind of technical solution.In the technical scheme, it is obtained from least one data source related to the reading context and scene of user
Search key include: to excavate and read with user from user's history log, internet web page and/or search log in real time
Read context and the relevant search key of scene.
Referring to Fig. 5, the method for pushing of described search keyword includes:
S51 excavates the reading with user or more from user's history log, internet web page and/or search log in real time
Text search key relevant with scene.
Described search keyword is excavated from user's history log, and it is crucial relevant search can be obtained according to historical data
Word.For example, user is after using search term " next stop " searching for Internet, often immediately according to the user's history log
Internet is searched again for using search term " subway ".At this point, by the excavation to user's history log, when user is in search box
When inputting " next stop ", search engine can recommend " subway " this search key.
Preferably, from user's history log, search key relevant to the reading context and scene of user is excavated
Include: that main body, point of interest and/or intent data are paid close attention to according to user, establishes the incidence matrix about user tag and scene；Root
According to incidence matrix described in user tag and scenario queries, described search keyword is obtained.
Described search keyword is excavated from internet web page, can be excavated according to the reading habit of user to target search
Keyword.For example, most of all includes " Dongcheng District " if user searches in the webpage browsed after " the Temple of Heaven " this search term
This keyword, then search engine can recommend " Dongcheng after user inputs " the Temple of Heaven " this search term in search box
This search key of area ".
Preferably, from internet web page, search key packet relevant to the reading context and scene of user is excavated
Include: according in the internet web page text or user to the click relationship of internet web page, dug from internet web page
Dig search key relevant to the reading context and scene of user.
Text in so-called internet web page refers in particular to user and uses the text in the webpage read in search engine process
This.Preferably, it can be realized by the participle to text in internet web page, the operation such as different degree calculating to search key
Excavation.
The click relationship of so-called internet web page, refer to performed by user in search sessions twice or
Relationship between the click action of web page interlinkage more than twice.It preferably, can be by jumping between different search terms
Jumping between the statistics or URL (Uniform resource locator, uniform resource locator) and search term of probability
The statistics of probability realizes the excavation to search key.
Described search keyword is excavated from real-time search log, the heat that can use other users searches word real-time synchronization
To the user for using search engine.For example, user is when using search engine, other most users all use " Anti-Japanese War Victory "
This heat searches word searching for Internet, then by the excavation to real-time search log, " Anti-Japanese War Victory " this heat can be searched word conduct
Search key, and described search keyword is pushed to the user using search engine.
S52 estimates its corresponding clicking rate to described search keyword, and is searched according to the clicking rate that estimation obtains to described
Rope keyword is ranked up.
S53 pushes described search keyword to the desktop terminal of user according to the order of the sequence.
The present embodiment is by excavating with user's from user's history log, internet web page and/or search log in real time
Context and the relevant search key of scene are read, realizes and described search keyword is obtained from least one data source
It takes, to realize the personalization of the search key of recommendation.
Fourth embodiment
The present embodiment further provides the method for pushing of search key based on the above embodiment of the present invention
A kind of technical solution of middle obtaining step.In the technical scheme, from user's history log, the reading with user or more is excavated
The relevant with scene search key of text includes: to pay close attention to main body, point of interest and/or intent data according to user, foundation about with
The incidence matrix of family label and scene；According to incidence matrix described in user tag and scenario queries, described search keyword is obtained.
Referring to Fig. 6, from user's history log, search key relevant to the reading context and scene of user is excavated
Include:
S61 pays close attention to main body, point of interest and/or intent data according to user, establishes the association about user tag and scene
Matrix.
The user pay close attention to main body refer to user when browsing the result of page searching that search engine provides, the main body of concern
Vocabulary.The point of interest refers to the word for representing interest of the user when browsing webpage.The intent data refers to that user uses
The intention of search engine progress internet hunt.It is understood that above-mentioned three kinds of data can pass through the browsing to user
The excavation of data and search engine usage log and obtain.Moreover, by the excavation to above-mentioned three kinds of data, can directly or
Indirectly obtain described search keyword.
In the present embodiment, to user group according to itself attribute distributing user label.For example, can be according to user's
Gender or age bracket distribute different user tags.Then, using the user tag as the horizontal axis coordinate of incidence matrix, and
Longitudinal axis label using the scene as incidence matrix is excavated according to paying close attention in main body, point of interest and/or intent data from user
Data are obtained, the incidence matrix is established.
S62 obtains described search keyword according to incidence matrix described in user tag and scenario queries.
It establishes after the incidence matrix, according to the horizontal axis label of incidence matrix described in the user tag and scene matching
With longitudinal axis label, so that the word that obtains the word stored in the incidence matrix, and will acquire is as described search keyword.
The present embodiment is established by paying close attention to main body, point of interest and/or intent data according to user about user tag and field
The incidence matrix of scape, and the incidence matrix according to user tag and scenario queries obtain described search keyword, thus real
Now excavated from the search key in user's history log.
5th embodiment
The present embodiment further provides the method for pushing of search key based on the above embodiment of the present invention
A kind of technical solution.In the technical scheme, it is obtained from least one data source related to the reading context and scene of user
Search key include: that relevant to the reading context and scene of user search key is excavated from internet web page.
Referring to Fig. 7, the method for pushing of described search keyword includes:
S71 excavates search key relevant to the reading context and scene of user from internet web page.
Specifically, excavating search key packet relevant to the reading context and scene of user from internet web page
Include: according in the internet web page text or user to the click relationship of internet web page, dug from internet web page
Dig search key relevant to the reading context and scene of user.
Illustratively, Fig. 8 is shown under a kind of preferred embodiment of the present embodiment, according in the internet web page
Text the flow chart for reading context and the relevant search key of scene to user is excavated from internet web page.Referring to
Fig. 8 is excavated from internet web page related to the reading context and scene of user according to the text in the internet web page
Search key include:
S81 inputs the webpage browsed after search term from user is obtained in user's history log.
S82 extracts representative keyword according to TF-IDF algorithm from the text of the webpage.
S83, using the representative keyword as described search keyword.
The TF-IDF algorithm is a kind of statistic algorithm, for assessing a words for the significance level of a file.
Preferably, the text that can first input in the webpage browsed after search term to user carries out word cutting, then to word cutting result point
Its different degree is not calculated using the TF-IDF algorithm, is selecting from different word cutting results importance sorting preferential, from
And realize the extraction to representative keyword.After completing to the extraction of representative keyword, it will extract
Representative keyword as described search keyword.
According to user to the click relationship of internet web page, excavated from internet web page with the reading context of user and
The relevant search key of scene includes: to jump relationship between search term in session according to user, is searched described in excavation
Rope keyword；Or relationship is jumped between URL and search term in a session according to user, it is crucial to excavate described search
Word.
It is understood that user may use different search term searching for Internet in a search sessions.And
The search key that front and back uses in search sessions has very big possibility in semanteme or there is logically associations.Therefore,
Relationship can be jumped between search terms different in a search sessions by excavating user, realized to described search keyword
Excavation.
In addition, user may access different webpages in a search sessions, may include in the URL of these webpages
With the presence of the associated word of search term used when being searched for user.It therefore, can be by excavating user in a search sessions
Relationship is jumped between URL and search term, excavates described search keyword.
It should be noted that there is the relationship that jumps between URL and search term can be user described in the use
After search term searching for Internet, the URL is clicked, can also be that user is clicking the URL, namely browsed described
After the corresponding webpage of URL, internet is searched again for using described search word.
Illustratively, Fig. 9 is shown under another preferred embodiment of the present embodiment, according to URL and search term it
Between jump relationship, excavate the process of described search keyword.Referring to Fig. 9, according to user in a session URL and search term
Between jump relationship, excavating described search keyword includes:
S91 is counted and is once jumped probability between URL and search term.
As it was noted above, user uses search term searching for Internet after clicking URL, or searched for mutually using search term
URL is clicked after networking, can be calculated as once jumping between URL and search term.
S92 once jumps jumping generally between probability calculation URL and search term according to described using Random Walk Algorithm
Rate.
Figure 10 shows the principle of the Random Walk Algorithm.Referring to Figure 10, calculating between URL101 and search term 102
When jumping probability, it is assumed that between another intermediate search word 103 and the URL101 and described search word 102 all exist jump
Relationship, then the intermediate search word 103 conditional transition probability phase between described search word 102 and the URL101 respectively
Multiply, and be directed to different intermediate search words 103, above-mentioned multiplied result is summed, the URL101 and described search have just been obtained
Probability is jumped between word 102.That is, the probability that jumps between the URL101 and described search word 102 can be by formula
(2) it provides:
S93 will jump the highest search term of probability as described search keyword.
Preferably, it can set and jump probability threshold value, and probability will be jumped and be higher than the search term for jumping probability threshold value
As described search keyword.The highest N number of search term of probability is jumped furthermore it is also possible to take, and using this N number of search term as institute
State search key.
S72 estimates its corresponding clicking rate to described search keyword, and is searched according to the clicking rate that estimation obtains to described
Rope keyword is ranked up.
S73 pushes described search keyword to the desktop terminal of user according to the order of the sequence.
The present embodiment is crucial by excavating search relevant to the reading context and scene of user from internet web page
Word, the search key for realizing the reading interest based on user excavates, and then realizes the personalization of search key recommendation.
Sixth embodiment
The present embodiment is based on the above embodiment of the present invention, in the method for pushing that further provides search key
A kind of technical solution of obtaining step.In the technical scheme, from real-time search log, the reading context with user is excavated
Search key relevant with scene includes: by excavating search log in real time, and statistics search term corresponds to retrieval amount, number of users,
With retrieval quantitative change rate；According to the retrieval amount of real-time statistics, number of users, and retrieval quantitative change rate, it is fitted searchable index function；Root
According to described search exponential function, search term popular on entire user group is selected, as described search keyword.
Referring to Figure 11, search key relevant to the reading context and scene of user is excavated from real-time search log
Include:
S111, by excavating search log in real time, statistics search term corresponds to retrieval amount, number of users, and retrieval quantitative change rate.
It is understood that each search term has its corresponding retrieval amount, number of users in the real-time search log.
In the present embodiment, by the excavation real-time search log, the retrieval amount of each search term, number of users are counted, and
The retrieval quantitative change rate of described search word is calculated by operation.
S112 is fitted searchable index function according to the retrieval amount of real-time statistics, number of users, and retrieval quantitative change rate.
Specifically, obtaining described search word pair by the retrieval amount, number of users and the calculating for retrieving quantitative change rate
The searchable index function answered.Further, described search exponential function is provided by formula (3):
Wherein, h (t) is searchable index function, and s (t) is the retrieval amount that statistics obtains, and n (t) is the user that statistics obtains
Number, rate (t) are retrieval quantitative change rates.
S113 selects search term, as described search keyword according to described search exponential function.
According to preset searchable index threshold value, described search keyword is chosen from numerous search terms.Preferably,
It can choose search term of the value of described search exponential function on preset searchable index minimum threshold, as institute
State search key.
The present embodiment corresponds to retrieval amount, number of users and the variation of retrieval amount by excavating search log, statistics search term in real time
Rate is fitted searchable index function, and search according to described according to the retrieval amount of real-time statistics, number of users, and retrieval quantitative change rate
Rope exponential function selects search term, as described search keyword, to realize crucial according to the search of search log in real time
Word excavates.
7th embodiment
The present embodiment is further provided and is searched in the method for pushing of plain keyword based on the above embodiment of the present invention
A kind of technical solution of obtaining step.In the technical scheme, according to described search exponential function, search term is selected, as
Before described search keyword, from real-time search log, excavates search relevant to the reading context and scene of user and close
Keyword further include: described search word is filtered；According to described search exponential function, search term is selected, is searched as described
After rope keyword, from real-time search log, search key relevant to the reading context and scene of user is excavated also
It include: to described search keyword, addition heat searches picture and/or heat searches voice；Picture searched to the heat and/or heat search voice into
Row compression, and store that compressed heat searches picture and/or heat searches voice.
Referring to Figure 12, search key relevant to the reading context and scene of user is excavated from real-time search log
Include:
S121, by excavating search log in real time, statistics search term corresponds to retrieval amount, number of users, and retrieval quantitative change rate.
S122 is fitted searchable index function according to the retrieval amount of real-time statistics, number of users, and retrieval quantitative change rate.
S123 is filtered described search word.
In order to guarantee the content health for the search key excavated, before selecting described search keyword, to alternative
Search term uniformly carry out information filtering.Specifically, can be according to pre-set deactivated vocabulary, to content in described search word
It is unsound to be filtered out.
S124 selects search term, as described search keyword according to described search exponential function.
S125, to described search keyword, addition heat searches picture and/or heat searches voice.
For the beauty and diversification for showing interface of described search keyword, for described search keyword add content with
Matched heat search picture.Furthermore it is also possible to which adding the matching heat of content for described search keyword searches voice.
S126 searches picture to the heat and/or heat is searched voice and compressed, and store compressed heat search picture and/or
Heat searches voice.
The present embodiment by according to described search exponential function select search term, and as described search keyword it
Before, described search word is filtered, and search term is being selected according to described search exponential function, and close as described search
After keyword, to described search keyword, addition heat searches picture and/or heat searches voice, and searches picture and/or heat to the heat
It searches voice to be compressed, and stores that compressed heat searches picture and/or heat searches voice, not only ensure that the search of push is crucial
Content health, and it is equipped with that heat searches picture and/or heat searches voice for described search keyword, so as to described search keyword
To show interface more various.
8th embodiment
The present embodiment is based on the above embodiment of the present invention, in the method for pushing that further provides search key
A kind of technical solution of sequence step.In the technical scheme, its corresponding clicking rate, and root are estimated to described search keyword
It includes: to estimate model estimation described search according to clicking rate that the clicking rate obtained according to estimates, which is ranked up described search keyword,
The clicking rate of keyword；Described search keyword is ranked up according to the clicking rate.
Referring to Figure 13, its corresponding clicking rate, and the clicking rate pair obtained according to estimation are estimated to described search keyword
Described search keyword, which is ranked up, includes:
S131 estimates the clicking rate of model estimation described search keyword according to clicking rate.
The clicking rate estimation model is provided by formula (1).Wherein, the feature vector of described search keyword passes through to described
The statistical analysis of the context of search key and obtain.Under normal conditions, the feature vector of described search keyword is and institute
State the higher one group of word of co-occurrence probabilities of search key.
S132 is ranked up described search keyword according to the clicking rate.
After completing the clicking rate estimation to described search keyword, according to the clicking rate that estimation obtains, searched to described
Rope keyword is ranked up.
The present embodiment is by estimating the clicking rate of described search keyword according to clicking rate model, and according to the click
Rate is ranked up described search keyword, to be distinguished to the validity of described search keyword, so as to described
The push of search key is more purposive.
9th embodiment
The present embodiment is based on the above embodiment of the present invention, in the method for pushing that further provides search key
A kind of technical solution of sequence step.In the technical scheme, estimating that model estimates described search keyword according to clicking rate
Clicking rate before, its corresponding clicking rate estimated to described search keyword, and the clicking rate obtained according to estimation is to described
Search key is ranked up further include: according to user's history behavior, obtains the feature vector of described search keyword；According to institute
Feature vector and target clicking rate are stated, the training clicking rate estimates model.
Referring to Figure 14, its corresponding clicking rate, and the clicking rate pair obtained according to estimation are estimated to described search keyword
Described search keyword, which is ranked up, includes:
S141 obtains the feature vector of described search keyword according to user's history behavior.
In the present embodiment, it is excavated according to feature vector of the historical behavior of user to described search keyword.It is excellent
Choosing, can excavate in a search sessions with the higher word of described search key words co-occurrence probability, as described search
Element in the feature vector of keyword.Element in described eigenvector can also be the corresponding page of described search keyword
Scene and/or user's scene.
It should be noted that the order of accuarcy of the characteristic vector data in order to guarantee the described search keyword got,
After the feature vector element for getting described search keyword according to co-occurrence probabilities, it is also necessary to the feature vector got
Element is once filtered.The purpose done so is to filter out insecure sample data in described eigenvector element.Tool
Body way can be to look at feature vector element in the co-occurrence in the historical behavior of all users between described search keyword
Probability.If described eigenvector element and the co-occurrence probabilities of described search keyword are inclined in the historical data of all users
It is low, then this feature vector element can be deleted from feature vector.
S142, according to described eigenvector and target clicking rate, the training clicking rate estimates model.
The clicking rate estimation model is provided by formula (1) hereinbefore.In the training stage of model, each training sample,
Namely search term all has its corresponding target clicking rate.According to the training sample and its target clicking rate, described in training
Clicking rate estimates model, until all weight vectors in clicking rate estimation model are restrained.
In addition, completing the clicking rate estimation model after training, it is also necessary to according to ROC (Receiver
Operating characteristic curve, Receiver operating curve) curve AUC (Area under curve,
Area under the curve) parameter and MAE (Mean absolute error, mean absolute error) estimate trained clicking rate
The predictive ability and prediction error of model are assessed.AUC is able to reflect the clicking rate estimation model and clicks behavior to user
The predictive ability of behavior is not clicked.MAE is able to reflect the size of the clicking rate estimation model predictive error.Using above-mentioned two
Kind of parameter assesses trained clicking rate estimation model, can guarantee the model clicking rate is estimated it is effective
Property and accuracy rate.
S143 estimates the clicking rate of model estimation described search keyword according to clicking rate.
S144 is ranked up described search keyword according to the clicking rate.
The present embodiment by according to clicking rate estimate model estimate described search keyword clicking rate before, according to
Family historical behavior obtains the feature vector of described search keyword, and according to described eigenvector and target clicking rate, training
The clicking rate estimates model, and the clicking rate estimation model is enabled accurately to provide the clicking rate of described search keyword
Estimated value.
Tenth embodiment
Present embodiments provide a kind of technical solution of the driving means of search key.Referring to Figure 15, in the technical side
In case, the driving means of described search keyword includes: to obtain module 151, sorting module 152 and pushing module 153.
The acquisition module 151 is used to obtain from least one data source relevant to the reading context and scene of user
Search key.
The sorting module 152 is obtained for estimating its corresponding clicking rate to described search keyword according to estimation
Clicking rate described search keyword is ranked up.
The pushing module 153 is used for crucial to the desktop terminal of user push described search according to the order of the sequence
Word.
Optionally, the pushing module 151 includes: layout designs unit and shows unit.
The layout designs unit is used for the screen width according to desktop terminal, designs the interface cloth of each search key
Office.
The unit that shows is for showing described search keyword, described search pass to user according to the interface layout
The corresponding heat of keyword searches picture and/or the corresponding heat of described search keyword searches voice.
Optionally, the acquisition module 151 is specifically used for: from user's history log, internet web page and/or search in real time
In log, search key relevant to the reading context and scene of user is excavated.
Optionally, when the Web log mining search key relevant to the reading context and scene of user from user's history
When, the acquisition module 151 includes: that matrix establishes unit and keyword acquiring unit.
The matrix establishes unit for paying close attention to the data such as main body, point of interest, intention according to user, establishes and marks about user
The incidence matrix of label and scene.
The keyword acquiring unit is used for the incidence matrix according to user tag and scenario queries, obtains described search
Keyword.
Optionally, when excavated from internet web page relevant to the reading context and scene of user search key it
When, it is described obtain module 151 be specifically used for: according in the internet web page text or user to internet web page
Click relationship excavates search key relevant to the reading context and scene of user from internet web page.
Optionally, when according to the text in the internet web page from excavated in internet web page with above and below the reading of user
When text search key relevant with scene, the acquisition module 151 includes: webpage acquiring unit, keyword extracting unit
And keyword determination unit.
The webpage acquiring unit is used to input the webpage browsed after search term from acquisition user in user's history log.
The keyword extracting unit is used to be extracted from the text of the webpage representative according to TF-IDF algorithm
Keyword.
The keyword determination unit is used for using the representative keyword as described search keyword.
Optionally, when according to user being excavated the click relationship of different internet web pages with user from internet web page
When reading context and the relevant search key of scene, the acquisition module 151 includes: that search term jumps relation excavation list
Member or URL jump relation excavation unit.
Described search word jumps relation excavation unit for the jump according to user between search terms different in a session
Transfer the registration of Party membership, etc. from one unit to another, excavates described search keyword.
The URL jump relation excavation unit for according to user in a session uniform resource position mark URL with search
Relationship is jumped between rope word, excavates described search keyword.
Optionally, the URL jumps relation excavation unit and is specifically used for: once jumping between statistics URL and search term
Probability；Using Random Walk Algorithm, probability is jumped between probability calculation URL and search term according to described once jump；It will jump
Turn the highest search term of probability as described search keyword.
Optionally, when the excavation search key relevant to the reading context and scene of user from real-time search log
When, the acquisition module 151 includes: statistic unit, fitting unit and Keyword Selection unit.
The statistic unit is used for by excavating search log in real time, and statistics search term corresponds to retrieval amount, number of users, and inspection
Rope amount change rate.
The fitting unit is used for according to the retrieval amount of real-time statistics, number of users, and retrieval quantitative change rate, and fitting search refers to
Number function.
The Keyword Selection unit is used to select search term according to described search exponential function, closes as described search
Keyword.
Optionally, when the excavation search key relevant to the reading context and scene of user from real-time search log
When, according to described search exponential function, select search term, before described search keyword, the acquisition module 151
Further include: filter element.
The filter element is for being filtered described search word.
Optionally, when the excavation search key relevant to the reading context and scene of user from real-time search log
When, search term, after described search keyword, the acquisition module 151 are being selected according to described search exponential function
Further include: multimedia adding unit and multimedia storing unit.
The multimedia adding unit is used for described search keyword, and addition heat searches picture and/or heat searches voice.
The multimedia storing unit is used to search picture to the heat and/or heat is searched voice and compressed, and stores compression
Heat afterwards searches picture and/or heat searches voice.
Optionally, the sorting module 152 includes: estimation unit and sequencing unit.
The estimation unit is used to estimate according to clicking rate the clicking rate of model estimation described search keyword, wherein institute
Clicking rate estimation model is stated to be given by:
The sequencing unit is for being ranked up described search keyword according to the clicking rate.
Optionally, the sorting module 152 further include: vector acquiring unit and model training unit.
The vector acquiring unit be used for according to clicking rate estimate model estimation described search keyword clicking rate it
Before, according to user's history behavior, obtain the feature vector of described search keyword.
The model training unit be used for according to clicking rate estimate model estimation described search keyword clicking rate it
Before, according to described eigenvector and target clicking rate, the training clicking rate estimates model.
Will be appreciated by those skilled in the art that each module of the above invention or each step can use general meter
Device is calculated to realize, they can be concentrated on single computing device, or be distributed in network constituted by multiple computing devices
On, optionally, they can be realized with the program code that computer installation can be performed, so as to be stored in storage
It is performed by computing device in device, perhaps they are fabricated to each integrated circuit modules or will be more in them
A module or step are fabricated to single integrated circuit module to realize.In this way, the present invention is not limited to any specific hardware and
The combination of software.
All the embodiments in this specification are described in a progressive manner, the highlights of each of the examples are with
The difference of other embodiments, the same or similar part between each embodiment may refer to each other.
The above description is only a preferred embodiment of the present invention, is not intended to restrict the invention, for those skilled in the art
For, the invention can have various changes and changes.All any modifications made within the spirit and principles of the present invention are equal
Replacement, improvement etc., should all be included in the protection scope of the present invention.
Claims (26)
1. a kind of method for pushing of search key characterized by comprising
Search key relevant to the reading context and scene of user is obtained from least one data source, wherein the use
The reading context at family refers to the context for the content that user is read by internet, and the scene includes page scene and user
Scene；
Its corresponding clicking rate is estimated to described search keyword, and the clicking rate obtained according to estimation is to described search keyword
It is ranked up；
Described search keyword is pushed to the desktop terminal of user according to the order of the sequence.
2. the method according to claim 1, wherein being pushed away according to the order of the sequence to the desktop terminal of user
The described search keyword is sent to include:
According to the screen width of desktop terminal, the interface layout of each search key is designed；
According to the interface layout, to user shows described search keyword, the corresponding heat of described search keyword searches picture and/
Or the corresponding heat of described search keyword searches voice.
3. the method according to claim 1, wherein obtaining the reading with user or more from least one data source
Relevant with the scene search key of text includes:
From user's history log, internet web page and/or search log in real time, the reading context and scene with user are excavated
Relevant search key.
4. according to the method described in claim 3, it is characterized in that, being excavated in the reading with user from user's history log
Hereafter relevant search key includes: with scene
Main body, point of interest and/or intent data are paid close attention to according to user, establishes the incidence matrix about user tag and scene；
According to incidence matrix described in user tag and scenario queries, described search keyword is obtained.
5. according to the method described in claim 3, it is characterized in that, excavating the reading with user or more from internet web page
Relevant with the scene search key of text includes:
According in the internet web page text or user to the click relationship of internet web page, from internet web page
Excavate search key relevant to the reading context and scene of user.
6. according to the method described in claim 5, it is characterized in that, according to the text in the internet web page, from internet
Search key relevant to the reading context and scene of user is excavated in webpage includes:
The webpage browsed after search term is inputted from user is obtained in user's history log；
According to TF-IDF algorithm, representative keyword is extracted from the text of the webpage；
Using the representative keyword as described search keyword.
7. according to the method described in claim 5, it is characterized in that, according to user to the click relationship of internet web page, from mutual
Search key relevant to the reading context and scene of user is excavated in intranet web includes:
Relationship is being jumped between search term in session according to user, is excavating described search keyword；Or
Relationship is jumped between uniform resource position mark URL and search term in a session according to user, excavates described search
Keyword.
8. the method according to the description of claim 7 is characterized in that according to user in a session between URL and search term
Jump relationship, excavating described search keyword includes:
Probability is once jumped between statistics URL and search term；
Using Random Walk Algorithm, probability is jumped between probability calculation URL and search term according to described once jump；
The highest search term of probability will be jumped as described search keyword.
9. according to the method described in claim 3, it is characterized in that, being excavated in the reading with user from real-time search log
Hereafter relevant search key includes: with scene
By excavating search log in real time, statistics search term corresponds to retrieval amount, number of users, and retrieval quantitative change rate；
According to the retrieval amount of real-time statistics, number of users, and retrieval quantitative change rate, it is fitted searchable index function；
According to described search exponential function, search term is selected, as described search keyword.
10. according to the method described in claim 9, it is characterized in that, select search term according to described search exponential function,
Before described search keyword, from real-time search log, search relevant to the reading context and scene of user is excavated
Rope keyword further include:
Described search word is filtered.
11. according to the method described in claim 9, it is characterized in that, select search term according to described search exponential function,
After described search keyword, from real-time search log, search relevant to the reading context and scene of user is excavated
Rope keyword further include:
To described search keyword, addition heat searches picture and/or heat searches voice；
Picture is searched to the heat and/or heat is searched voice and compressed, and stores that compressed heat searches picture and/or heat searches voice.
12. the method according to claim 1, wherein estimate its corresponding clicking rate to described search keyword,
And described search keyword is ranked up according to the clicking rate that estimation obtains and includes:
The clicking rate of model estimation described search keyword is estimated according to clicking rate, wherein the clicking rate estimation model is under
Formula provides:
Wherein, CTR indicates the clicking rate of search key, CTRtemplateIndicate the click of the search key in advertisement position stage
Rate, CTRluIndicate the clicking rate of the search key in link unit stage；xiIndicate advertisement position stage described search keyword,
The feature vector element of corresponding page scene, user's scene, wiIndicate the corresponding weight vectors member of described eigenvector element
Element；xjIt indicates in connection unit stage described search keyword, the feature vector element of corresponding page scene, user's scene, wj
It indicates in the corresponding weight vectors element of link unit stage described eigenvector element；
Described search keyword is ranked up according to the clicking rate.
13. according to the method for claim 12, which is characterized in that estimating that model estimation described search is closed according to clicking rate
Before the clicking rate of keyword, its corresponding clicking rate, and the clicking rate pair obtained according to estimation are estimated to described search keyword
Described search keyword is ranked up further include:
According to user's history behavior, the feature vector of described search keyword is obtained；
According to described eigenvector and target clicking rate, the training clicking rate estimates model.
14. a kind of driving means of search key characterized by comprising
Module is obtained, it is crucial for obtaining search relevant to the reading context and scene of user from least one data source
Word, wherein the reading context of the user refers to the context for the content that user is read by internet, and the scene includes
Page scene and user's scene；
Sorting module, for estimating its corresponding clicking rate, and the clicking rate pair obtained according to estimation to described search keyword
Described search keyword is ranked up；
Pushing module, for pushing described search keyword to the desktop terminal of user according to the order of the sequence.
15. device according to claim 14, which is characterized in that the pushing module includes:
Layout designs unit designs the interface layout of each search key for the screen width according to desktop terminal；
Show unit, for it is corresponding to show described search keyword, described search keyword to user according to the interface layout
Heat search picture and/or the corresponding heat of described search keyword searches voice.
16. device according to claim 14, which is characterized in that the acquisition module is specifically used for:
From user's history log, internet web page and/or search log in real time, the reading context and scene with user are excavated
Relevant search key.
17. device according to claim 16, which is characterized in that when the reading of the Web log mining from user's history and user
When context and the relevant search key of scene, the acquisition module includes:
Matrix establishes unit, for paying close attention to main body, point of interest, intent data according to user, establishes about user tag and scene
Incidence matrix；
Keyword acquiring unit is used for the incidence matrix according to user tag and scenario queries, obtains described search keyword.
18. device according to claim 16, which is characterized in that excavated in the reading with user when from internet web page
Hereafter with scene when relevant search key, the acquisition module is specifically used for:
According in the internet web page text or user to the click relationship of internet web page, from internet web page
Excavate search key relevant to the reading context and scene of user.
19. device according to claim 18, which is characterized in that when according to the text in the internet web page from interconnection
When excavating search key relevant to the reading context and scene of user in net webpage, the acquisition module includes:
Webpage acquiring unit, for inputting the webpage browsed after search term from acquisition user in user's history log；
Keyword extracting unit, for extracting representative key from the text of the webpage according to TF-IDF algorithm
Word；
Keyword determination unit, for using the representative keyword as described search keyword.
20. device according to claim 18, which is characterized in that when the click pass according to user to different internet web pages
When system excavates search key relevant to the reading context and scene of user from internet web page, the acquisition module
Include:
Search term jumps relation excavation unit, for jumping relationship between search terms different in a session according to user,
Excavate described search keyword；Or
URL jumps relation excavation unit, for according to user in a session between uniform resource position mark URL and search term
Jump relationship, excavate described search keyword.
21. device according to claim 20, which is characterized in that the URL jumps relation excavation unit and is specifically used for:
Probability is once jumped between statistics URL and search term；
Using Random Walk Algorithm, probability is jumped between probability calculation URL and search term according to described once jump；
The highest search term of probability will be jumped as described search keyword.
22. device according to claim 16, which is characterized in that when the reading of excavation and user from real-time search log
When context and the relevant search key of scene, the acquisition module includes:
Statistic unit, for by excavating search log in real time, statistics search term to correspond to retrieval amount, number of users, and retrieval quantitative change
Rate；
Fitting unit, for being fitted searchable index function according to the retrieval amount of real-time statistics, number of users, and retrieval quantitative change rate；
Keyword Selection unit, for search term being selected, as described search keyword according to described search exponential function.
23. device according to claim 22, which is characterized in that when the reading of excavation and user from real-time search log
When context and the relevant search key of scene, according to described search exponential function, search term is selected, is searched as described
Before rope keyword, the acquisition module further include:
Filter element, for being filtered to described search word.
24. device according to claim 22, which is characterized in that when the reading of excavation and user from real-time search log
When context and the relevant search key of scene, search term is being selected according to described search exponential function, is being searched as described
After rope keyword, the acquisition module further include:
Multimedia adding unit, for adding to described search keyword, heat searches picture and/or heat searches voice；
Multimedia storing unit for searching picture to the heat and/or heat is searched voice and compressed, and stores compressed heat and searches
Picture and/or heat search voice.
25. device according to claim 14, which is characterized in that the sorting module includes:
Estimation unit, for estimating the clicking rate of model estimation described search keyword according to clicking rate, wherein the clicking rate
Estimation model is given by:
Wherein, CTR indicates the clicking rate of search key, CTRtemplateIndicate the click of the search key in advertisement position stage
Rate, CTRluIndicate the clicking rate of the search key in link unit stage；xiIndicate advertisement position stage described search keyword,
The feature vector element of corresponding page scene, user's scene, wiIndicate the corresponding weight vectors member of described eigenvector element
Element；xjIt indicates in connection unit stage described search keyword, the feature vector element of corresponding page scene, user's scene, wj
It indicates in the corresponding weight vectors element of link unit stage described eigenvector element；
Sequencing unit, for being ranked up according to the clicking rate to described search keyword.
26. device according to claim 25, which is characterized in that the sorting module further include:
Vector acquiring unit, for according to clicking rate estimate model estimation described search keyword clicking rate before, according to
User's history behavior obtains the feature vector of described search keyword；
Model training unit, for according to clicking rate estimate model estimation described search keyword clicking rate before, according to
Described eigenvector and target clicking rate, the training clicking rate estimate model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510475976.9A CN105045901B (en) | 2015-08-05 | 2015-08-05 | The method for pushing and device of search key |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510475976.9A CN105045901B (en) | 2015-08-05 | 2015-08-05 | The method for pushing and device of search key |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105045901A CN105045901A (en) | 2015-11-11 |
| CN105045901B true CN105045901B (en) | 2019-04-30 |
Family
ID=54452448
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510475976.9A Active CN105045901B (en) | 2015-08-05 | 2015-08-05 | The method for pushing and device of search key |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105045901B (en) |
Families Citing this family (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105488027B (en) * | 2015-11-30 | 2019-07-12 | 百度在线网络技术（北京）有限公司 | The method for pushing and device of keyword |
| CN106708282B (en) | 2015-12-02 | 2019-03-19 | 北京搜狗科技发展有限公司 | A kind of recommended method and device, a kind of device for recommendation |
| CN105678335B (en) * | 2016-01-08 | 2019-07-02 | 车智互联（北京）科技有限公司 | It estimates the method, apparatus of clicking rate and calculates equipment |
| CN107562750A (en) * | 2016-06-30 | 2018-01-09 | 百度在线网络技术（北京）有限公司 | A kind of method and apparatus for providing search result |
| CN106844484B (en) * | 2016-12-23 | 2020-08-28 | 北京安云世纪科技有限公司 | Information searching method and device and mobile terminal |
| CN106919641B (en) * | 2017-01-12 | 2020-04-17 | 北京三快在线科技有限公司 | Interest point searching method and device and electronic equipment |
| CN106919693B (en) * | 2017-03-07 | 2020-12-01 | 阿里巴巴（中国）有限公司 | Method and device for improving hot word exposure coverage rate |
| CN107463704B (en) * | 2017-08-16 | 2021-05-07 | 北京百度网讯科技有限公司 | Search method and device based on artificial intelligence |
| CN110020045B (en) * | 2017-09-25 | 2021-07-27 | 北京国双科技有限公司 | Keyword path analysis method and device |
| CN110069542B (en) * | 2017-09-26 | 2021-06-29 | 北京国双科技有限公司 | Keyword evaluation method and device |
| CN107885856A (en) * | 2017-11-16 | 2018-04-06 | 阿里巴巴集团控股有限公司 | A kind of page display method and device |
| CN108197244A (en) * | 2017-12-29 | 2018-06-22 | 北京奇虎科技有限公司 | It is a kind of to search for the method for pushing and device for recommending word |
| CN108460104B (en) * | 2018-02-06 | 2021-06-18 | 北京奇虎科技有限公司 | Method and device for customizing content |
| CN109033411A (en) * | 2018-08-06 | 2018-12-18 | 深圳市嘀哒知经科技有限责任公司 | A kind of searching method of knowledge expertise |
| CN109710088B (en) * | 2018-12-29 | 2022-12-27 | 北京金山安全软件有限公司 | Information searching method and device |
| CN109636491A (en) * | 2019-01-25 | 2019-04-16 | 西窗科技(苏州)有限公司 | A kind of optimization method and device that search engine advertisement keyword is launched |
| CN110674406A (en) * | 2019-09-29 | 2020-01-10 | 百度在线网络技术（北京）有限公司 | Recommendation method and device, electronic equipment and storage medium |
| CN113360773B (en) * | 2021-07-07 | 2023-07-04 | 脸萌有限公司 | Recommendation method and device, storage medium and electronic equipment |
| CN113792136A (en) * | 2021-08-25 | 2021-12-14 | 北京库睿科技有限公司 | Text data diversified recommendation type searching method and system |
| CN114491253B (en) * | 2022-01-21 | 2023-09-26 | 北京百度网讯科技有限公司 | Method and device for processing observation information, electronic equipment and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103064853A (en) * | 2011-10-20 | 2013-04-24 | 北京百度网讯科技有限公司 | Search suggestion generation method, device and system |
| CN103092877A (en) * | 2011-11-04 | 2013-05-08 | 百度在线网络技术（北京）有限公司 | Method and device for recommending keyword |
| CN103164521A (en) * | 2013-03-11 | 2013-06-19 | 亿赞普(北京)科技有限公司 | Keyword calculation method and device based on user browse and search actions |
| CN103365839A (en) * | 2012-03-26 | 2013-10-23 | 腾讯科技（深圳）有限公司 | Recommendation search method and device for search engines |
| CN103729351A (en) * | 2012-10-10 | 2014-04-16 | 阿里巴巴集团控股有限公司 | Search term recommendation method and device |
| CN104598583A (en) * | 2015-01-14 | 2015-05-06 | 百度在线网络技术（北京）有限公司 | Method and device for generating query sentence recommendation list |
| CN104750873A (en) * | 2015-04-22 | 2015-07-01 | 百度在线网络技术（北京）有限公司 | Popular search term push method and device |
-
2015
- 2015-08-05 CN CN201510475976.9A patent/CN105045901B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103064853A (en) * | 2011-10-20 | 2013-04-24 | 北京百度网讯科技有限公司 | Search suggestion generation method, device and system |
| CN103092877A (en) * | 2011-11-04 | 2013-05-08 | 百度在线网络技术（北京）有限公司 | Method and device for recommending keyword |
| CN103365839A (en) * | 2012-03-26 | 2013-10-23 | 腾讯科技（深圳）有限公司 | Recommendation search method and device for search engines |
| CN103729351A (en) * | 2012-10-10 | 2014-04-16 | 阿里巴巴集团控股有限公司 | Search term recommendation method and device |
| CN103164521A (en) * | 2013-03-11 | 2013-06-19 | 亿赞普(北京)科技有限公司 | Keyword calculation method and device based on user browse and search actions |
| CN104598583A (en) * | 2015-01-14 | 2015-05-06 | 百度在线网络技术（北京）有限公司 | Method and device for generating query sentence recommendation list |
| CN104750873A (en) * | 2015-04-22 | 2015-07-01 | 百度在线网络技术（北京）有限公司 | Popular search term push method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105045901A (en) | 2015-11-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105045901B (en) | The method for pushing and device of search key | |
| CN105159930B (en) | The method for pushing and device of search key | |
| CN102542003B (en) | For taking the click model of the user view when user proposes inquiry in a search engine into account | |
| KR101667344B1 (en) | Method and system for providing search results | |
| WO2017190610A1 (en) | Target user orientation method and device, and computer storage medium | |
| US9798820B1 (en) | Classification of keywords | |
| CN107424043A (en) | A kind of Products Show method and device, electronic equipment | |
| US20110320429A1 (en) | Systems and methods for augmenting a keyword of a web page with video content | |
| US20130054672A1 (en) | Systems and methods for contextualizing a toolbar | |
| US20120130969A1 (en) | Generating context information for a search session | |
| Paranjape et al. | Improving website hyperlink structure using server logs | |
| US20120290974A1 (en) | Systems and methods for providing a discover prompt to augmented content of a web page | |
| JP2015537259A (en) | Ranking search results based on click-through rate | |
| CN104933239A (en) | Hybrid model based personalized position information recommendation system and realization method therefor | |
| TW201007488A (en) | System and method for trail identification with search results | |
| WO2011146874A2 (en) | Microsite models | |
| US20110236870A1 (en) | System and method for learning | |
| CN101650731A (en) | Method for generating suggested keywords of sponsored search advertisement based on user feedback | |
| CN103064945A (en) | Situation searching method based on body | |
| WO2012051470A1 (en) | Systems and methods for using a behavior history of a user to augment content of a webpage | |
| US20100010982A1 (en) | Web content characterization based on semantic folksonomies associated with user generated content | |
| CN107592930A (en) | The automatic of application crawls | |
| US20120246561A1 (en) | Systems and methods for extended content harvesting for contextualizing | |
| US20140280133A1 (en) | Structured Data to Aggregate Analytics | |
| US9104746B1 (en) | Identifying contrarian terms based on website content |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |